Reciprocal regulation of endothelial–mesenchymal transition by MAPK7 and EZH2 in intimal hyperplasia and coronary artery disease

Endothelial–mesenchymal transition (EndMT) is a form of endothelial dysfunction wherein endothelial cells acquire a mesenchymal phenotype and lose endothelial functions, which contributes to the pathogenesis of intimal hyperplasia and atherosclerosis. The mitogen activated protein kinase 7 (MAPK7) inhibits EndMT and decreases the expression of the histone methyltransferase Enhancer-of-Zeste homologue 2 (EZH2), thereby maintaining endothelial quiescence. EZH2 is the catalytic subunit of the Polycomb Repressive Complex 2 that methylates lysine 27 on histone 3 (H3K27me3). It is elusive how the crosstalk between MAPK7 and EZH2 is regulated in the endothelium and if the balance between MAPK7 and EZH2 is disturbed in vascular disease. In human coronary artery disease, we assessed the expression levels of MAPK7 and EZH2 and found that with increasing intima/media thickness ratio, MAPK7 expression decreased, whereas EZH2 expression increased. In vitro, MAPK7 activation decreased EZH2 expression, whereas endothelial cells deficient of EZH2 had increased MAPK7 activity. MAPK7 activation results in increased expression of microRNA (miR)-101, a repressor of EZH2. This loss of EZH2 in turn results in the increased expression of the miR-200 family, culminating in decreased expression of the dual-specificity phosphatases 1 and 6 who may repress MAPK7 activity. Transfection of endothelial cells with miR-200 family members decreased the endothelial sensitivity to TGFβ1-induced EndMT. In endothelial cells there is reciprocity between MAPK7 signaling and EZH2 expression and disturbances in this reciprocal signaling associate with the induction of EndMT and severity of human coronary artery disease.

EZH2 regulates DUSP-1 and DUSP-6 expression through miR200a-c. EZH2 expression determines the level of MAPK7 activity. However, EZH2 is a transcriptional repressor that cannot directly regulate the activity of a kinase. MAPK7 activity is regulated by the Dual Specificity Phosphatases (DUSP)-1 and DUSP-6 21 , Figure 1. Reciprocity between MAPK7 and EZH2 in human coronary artery disease. (a-c) Representative pictures of Verhoeff-stained human coronary artery samples (n = 4-8) with increasing IMT, IMT < 1 (a) IMT 1-3 (b) and IMT > 3 (c). Intima-media thickness was measured (µm intima /µm media ) and samples stratified into three groups based on their intima-media thickness (d). An increasing IMT coincides with a progressively decreasing lumen area of the coronary artery (e), suggestive of progressive stenosis. MAPK7 expression levels were determined by qPCR and normalized to IMT < 1 (f). MAPK7 decreases with increasing IMT (g). EZH2 expression levels were determined by qPCR and normalized to IMT < 1 (h). EZH2 expression increased with increasing IMT (i). Data is expressed as mean ± S.D. of all individual observations. Statistical analysis was performed by ANOVA followed by Bonferroni post hoc tests. Correlations were performed using Pearson correlation. *p < 0.05, ***p < 0.001.

Figure 2.
Reciprocal signaling between MAPK7 and EZH2 in endothelial cells. EZH2 expression levels were determined by qPCR in HUVEC exposed to FSS (20 dyne/cm 2 ) compared to static controls (a). EZH2 protein levels were determined by western blot in HUVEC exposed to FSS and compared to static control (b). MAPK7 expression levels were determined by qPCR in HUVEC exposed to FSS, and HUVEC that are deficient in EZH2 (shEZH2) (c). MAPK7 activation (pMAPK7) levels were determined by immunoblotting and normalized to total MAPK7 protein levels (d). Protein expression of EZH2 and MAPK7 activation were associated in endothelial cells (e). Data is expressed as mean ± S.D. of all individual observations. Gene and protein expression data were obtained from 4 and 3 independent experiments, respectively. Comparisons between 2 groups were performed by Student t-tests and data from multiple groups were analyzed by ANOVA followed by Bonferroni post hoc tests. Correlations were performed using Pearson correlation. ***p < 0.001. www.nature.com/scientificreports/ yet a reduction in EZH2 expression is associated with a decreased expression of DUSP-1 and DUSP-6 23,24 . Therefore, we investigated alternative mechanisms that might decrease DUSP expression upon the reduction of EZH2. In silico analysis, using Targetscan.org 25 Fig. 4a-e). Moreover, knockdown of EZH2 similarly increased the expression of microRNAs in these clusters ( Fig. 4a-e). The EZH2-induced trimethylation of lysine 27 on histone 3 (H3K27Me3) is present in the promoter regions of the miR-200b/a/429 and miR-200c/141 clusters (Fig. 4f,i). Knockdown of EZH2 in endothelial cells reduced the level of H3K27Me3 at these gene regions (Fig. 4f,i), and the loss of this repressive histone mark coincided with increased expression of miR-200a-c, miR-141 and miR-429. In endothelial cells with constitutively active MAPK7 signaling (MEK5D), the abundance of H3K27Me3 is decreased at the promoter regions of miR-200b/a/429 (1.6-fold, p = 0.034; Fig. 4g,h) and miR-200c/141 (1.9-fold, p = 0.035; Fig. 4j,k), implying that MAPK7 activity increases the expression of miR-200 family members through a decrease in EZH2-mediated gene silencing.

Scientific
In luciferase reporter assays, all miR-200 family members were able to bind to the 3'UTR of DUSP-1 (Fig. 5a), but only miR-200a and miR-141 were able to bind the 3'UTR of DUSP-6 ( Fig. 5c). Corroboratively, exogenous expression of all miR-200 family members in endothelial cells decreased DUSP-1 expression (Fig. 5b), whereas only miR-200a and miR-141 decreased the expression of DUSP-6 ( Fig. 5d). Collectively, these data imply that the activation of MAPK7 by uniform LSS decreases the expression of DUSP-1 and DUSP-6 expression via the EZH2-dependent regulation of miR-200b/a/429 and miR-200c/141 expression. MiR-101 expression levels were determined by qPCR in HUVEC exposed to FSS (20 dyne/cm 2 ) with or without the MAPK7 inhibitor BIX02189 (10 µM) and normalized to the level of static controls (a). Luciferase reporter binding assays were performed for the 3'UTR of EZH2 in COS7 cells with ectopic expression of miR-101 or scrambled control sequences (scr). Luciferase activity was normalized to non-transfected cells (b). EZH2 and MAPK7 expression levels were determined by qPCR in HUVEC with ectopic expression of miR-101 or SCR and normalized to control (c,e). EZH2 and MAPK7 protein levels were determined by western blot in HUVEC with ectopic expression of miR-101 or scrambled control sequences (d,f). MiR-101 expression levels were determined by qPCR and normalized to IMT < 1 (g). MiR-101 decreases with increasing IMT (h), and associates with MAPK7 (i), and tends to associate with EZH2 (j) expression levels. In coronary artery disease, MAPK7 expression negatively correlates to EZH2 expression (k). Data is expressed as mean ± S.D. of all individual observations. In vitro experimental data was derived from 4 independent experiments, whereas human in vivo data was derived from n = 4-8 samples per group. Comparisons between 2 groups were performed by Student t-tests and data from multiple groups were analyzed by ANOVA followed by Bonferroni post hoc tests. Correlations were performed using Pearson correlation. *p < 0.05, **p < 0.01, ***p < 0.001.  Fig. 6a; Suppl. Fig. 3), albeit not statistically significant (p = 0.202), and did not affect EZH2 expression ( Fig. 6b; Suppl. Fig. 3). The addition of Simvastatin -a known activator of MAPK7 signaling 15 -did increase increased MAPK7 activation (5.8-fold, p < 0.001; Fig. 6a; Suppl. Fig. 3) and decreased EZH2 protein expression (3.0-fold, p < 0.001; Fig. 6b; Suppl. Fig. 3). The addition of BCI to simvastatin-treated endothelial cells did not increase the levels of MAPK7 activation nor decrease the protein expression of EZH2 further. Rather, the addition of BCI reduced the simvastatin-induced activity of MAPK by ~ 27% (p = 0.033; Fig. 6a; Suppl. Fig. 3).

Scientific
In human coronary artery disease, DUSP-1 expression is increased in advanced lesions (IMT > 3, p < 0.001, Fig. 6c) and increasing IMT associates with increased DUSP-1 expression (r 2 = 0.2767, p = 0.0301; Fig. 6d). Moreover, the increase in DUSP-1 expression associates with increased EZH2 expression in coronary artery disease (r 2 = 0.4541, p = 0.0030; Fig. 6e) and the increase in DUSP-1 expression tends to associate with decreased MAPK7 expression (r 2 = 0.1686, p = 0.1016; Fig. 6f), although not significantly. Also, DUSP-6 seems to be increased in coronary artery disease (p < 0.1, Fig. 6g), albeit not significantly. The apparent increase in DUSP-6 expression does not significantly associate with increasing IMT (r 2 = 0.0681, p = 0.3116; Fig. 6h and miR-429 (e) expression levels were determined by qPCR in HUVEC exposed to FSS (20 dyne/cm 2 ), transduced HUVEC with scr or shEZH2 and normalized to static control cells. H3K27me3 enrichment around the transcription start site (TSS) of the miR-200b/a/429 cluster was determined using ChIP in shEZH2-HUVEC and scr-HUVEC (f) and MEK5D-HUVEC and EV-HUVEC (g). H3K27me3 enrichment is shown as area under the curve (AUC) compared to input samples (h). H3K27me3 enrichment at the around the TSS of miR-200c/141 cluster was determined using ChIP in scr-HUVEC and shEZH2 (i) and in MEK5D-HUVEC and EV-HUVEC (j). H3K27me3 enrichment is shown as AUC compared to input samples (k). Data is expressed as mean ± S.D. of all individual observations. Gene expression data were obtained from 4 independent experiments. Data from ChIP experiments were obtained from three independent experiments. All data was analyzed by ANOVA followed by Bonferroni post hoc tests. *P < 0.05, **P < 0.01, ***P < 0.001.

Discussion
In this study, we show that reciprocity exists between the atheroprotective MAPK7 activation and the expression of histone methyltransferase EZH2 in endothelial cells. The reciprocity is regulated by the MAPK7-induced silencing of EZH2 expression by miR-101 and the EZH2-mediated silencing of the miR-200 family, which increases DUSP-1 and DUSP-6 expression and inhibits MAPK7 activation (Fig. 8). The reciprocity between MAPK7-EZH2 might reflect an autoregulatory feedback loop in endothelial cells that ensures endothelial homeostasis. As such, disturbances in this reciprocity leading to increased EZH2 expression can induce endothelial dysfunction and EndMT. In contrary artery disease-a condition associated with EndMT 1,26 -the reciprocity between MAPK7 and EZH2 is disturbed, resulting in elevated expression of DUSP-1 and EZH2 and the decreased expression of MAPK7. Restoring the reciprocity by ectopic expression of miR-101/200a/429 precludes EndMT and might offer therapeutic benefit in coronary artery disease. Data is expressed as mean ± S.D. of all individual observations. Gene expression data and Luciferase activity measurements were obtained from 4 independent experiments. All data was analyzed by ANOVA followed by Bonferroni post hoc tests. *P < 0.05, **P < 0.01, ***P < 0.001. www.nature.com/scientificreports/ EndMT contributes to intimal hyperplasia during coronary artery disease 1-5 , wherein MAPK7 signaling plays a protective role 5,16 . EndMT can be induced by hypoxia, inflammatory and fibrogenic signaling 6,27 . Transforming growth factor beta (TGFβ) induces EndMT canonically through the activation of downstream mediators Smad2/3, which culminates in the activation of the EndMT transcription factors Snail, Slug and Twist1 6 . MAPK7 inhibits EndMT 5,28 , potentially via the increased expression of inhibitory Smad7 29,30 and ID proteins 31 or the repression of TGF control elements in the promoter region of mesenchymal genes 32 . Yet, during intimal hyperplasia the signaling activity of MAPK7 is rapidly lost 28 .

Scientific Reports
DUSP-1 and DUSP-6 expression levels are elevated in a number of cardiovascular diseases and DUSP-1 deficient mice are protected from atherosclerosis development 33,34 . The elevated expression of DUSPs might explain the loss in protective MAPK7 signaling activity during coronary artery disease. Yet we could not substantiate this Figure 6. Inhibition of DUSP activity by BCI does not affect MAPK7 activity or EZH2 expression. MAPK7 (a) and EZH2 (b) protein expressions were determined using western blotting, in HUVEC treated with 5 µM of the DUSP-1/6 small molecule inhibitor BCI or co-treated with BCI and 10 µM Simvastatin for 24 h. Data was normalized to vehicle control cells. DUSP-1 expression levels were determined in human coronary artery disease samples by qPCR and normalized to IMT < 1 (c). DUSP-1 expression increases with increasing IMT (d) and is associated to EZH2 expression levels (e). DUSP-1 levels tend to show a negative correlation with MAPK7 levels in coronary artery disease (f). DUSP-6 expression levels were determined by qPCR and normalized to IMT < 1 (g). DUSP-6 expression is elevated in coronary artery disease but does not associate to the severity of disease (h), the level of EZH2 expression (i), nor the level of MAPK7 expression (j). Sim = Simvastatin (10 µM). Data is expressed as mean ± S.D. of all individual observations. Cell culture data was obtained from 4 independent experiments, whereas human in vivo data was derived from n = 4-8 samples per group. Data from multiple groups were analyzed by ANOVA followed by Bonferroni post hoc tests. Correlations were performed using Pearson correlation. **p < 0.01, ***p < 0.001. www.nature.com/scientificreports/ www.nature.com/scientificreports/ hypothesis as pharmacological inhibition of DUSP-1/6 in our experiments did not activate MAPK7 signaling, which may also depend on the availability of ATP and scaffolding proteins, and the activity of upstream MAPK kinases 35 . Besides, the regulation of MAPK7 activity by DUSPs is more complex than direct inactivation. Indeed, DUSP activity is greatly affected by posttranslational modifications [36][37][38][39][40] . For instance, DUSP activity is enhanced by its acetylation 37,38 or phosphorylation 36 , whereas its oxidation reduces its activity 39,40 . In our experiments, we did not investigate the posttranslational modifications of DUSP1, and in the absence of a response to the small molecule BCI, it is tempting to speculate that the prerequisites for DUSP1 activity were not met in our cell culture model. Moreover, the family of DUSPs contains over twenty members with overlapping substrate specificity 41 , that are not all inhibited by the addition of BCI 42 . It is therefore conceivable that alternative DUSPs maintain the inhibition of MAPK7 activity in the presence of BCI as a compensatory mechanism. Nonetheless, the expression of DUSP-1 is associated with an increasing IMT, and a decreased expression of MAPK7 in coronary artery disease. The expression of DUSP-1 and -6 is associated with high expression of EZH2 in various oncology's 23,24 , albeit by a currently unknown mechanism. We found that EZH2 silences the expression of the microRNA-200 family, which posttranscriptionally regulate the expression of DUSP-1 and DUSP-6. The loss of EZH2 expression by fluid shear stress therefore might increase the expression of miR-200 family members and decrease the expression of the DUSPs culminating in atheroprotective MAPK7 activation. Interestingly, the endothelial cell-specific overexpression of miR-200b precludes EndMT and alleviates diabetic cardiomyopathy in mice 43 . In coronary artery disease, EZH2 expression levels are elevated and high EZH2 expression is associated with endothelial dysfunction 18,19 . In combination, our current data might explain these observations and unifies them into a single mechanism, linking endothelial mechanotransduction to the epigenetic regulation of MAPK7 activity, potentially through DUSP-1 and DUSP-6. This double negative feedback loop might resemble a sensitive autoregulatory mechanism that ensures endothelial homeostasis, which when disturbed culminates in EndMT and possibly coronary artery disease.

Scientific Reports
It should be noted here, that the downstream effects of EZH2 and MAPK7 on atherogenesis may be much broader than the reciprocal regulation of their expression and activation. Indeed, in earlier work we used RNA sequencing and found that high EZH2 expression coincides with endothelial cell proliferation and www.nature.com/scientificreports/ RNAi-mediated silencing of EZH2 results in endothelial quiescence 18 . Using a similar approach, others reported that high EHZ2 associates with a reduced angiogenic potential and inflammatory activation of endothelial cells 19 . Similarly, the endothelial-specific genetic deletion of MAPK7 aggravates atherogenesis by the induction of endothelial cell inflammatory activation and loss of atheroprotective NO synthesis 16 and the induction of EndMT 5 . Thus, in future perspective it would be of high interest to validate and extend our current findings in atherosclerosis-prone mouse models (e.g. APOE-deficient or LDLR-deficient mice on a cholesterol-rich diet) wherein MAPK7 and EZH2 expression can be perturbed specifically in the endothelium. Such studies would benefit from the use of transcriptomic, proteomic or "multi-omics" approaches to unravel the multitude of targets and the principle downstream pathways EZH2 and MAPK7 regulate in the atherosclerotic endothelium.

Study limitations.
We acknowledge that our study is not without limitations. First, we only included unique 17 coronary artery samples from 10 subjects, which may be a limited number if confounders as hypertension, diabetes or smoking status are to be analyzed. In our study, we chose to stratify these coronary artery samples based on IMT (as a surrogate marker of disease severity) to investigate if the expression of MAPK and EZH2 are "disease-state" specific, and thereby ignored interindividual variation, which may be caused by common risk factors 44,45 or genetic susceptibility 46,47 . In our study, mean age, hypertension, diabetes and smoking status did not differ between stratified groups and no subject was overrepresented in any group. Using this approach, we show that there is a disbalance in MAPK7 activity and EZH2 expression in coronary artery disease and that this disbalance is perturbed with increasing IMT. In future perspective, it would be highly interesting to investigate how common risk factors affect the disbalance between MAPK7 activity and EZH2 expression in larger cohort studies. In addition, and since polymorphisms in MAPK7 48-50 and EZH2 51-54 are identified, it would be highly interesting to investigate whether these polymorphisms associate to a higher cardiovascular risk. Second, transcriptional expression was performed on of whole artery sections, meaning that the observed differences in MAPK7 and EZH2 expression are not necessarily derived from endothelial cells, but may reflect a significant change in other cell numbers, e.g. inflammatory and smooth muscle cells. Indeed, the cellular composition of a coronary artery lesion is dynamic and may change with disease progression 55,56 . Although MAPK7 expression appears endothelial cell-restricted in porcine and mouse arteries 28 , EZH2 is expressed by a plethora of cells. In vivo validation of our current findings by localization techniques such as double immunofluorescence would have to be performed in order to show endothelial specificity of the reported results. Moreover, the endothelial cell-specific deletion of MAPK7 and EZH2 in atherosclerosis-prone mice would further emphasize their relevance to atherogenesis in future perspective.
Third, for the mechanistic experiments detailed here, we used HUVEC rather than primary human coronary artery endothelial cells (HCAEC), and their distinct origin might interfere on our observations. Although the endothelium harbors a larger heterogeneity in vivo between endothelial cells from distinct vascular beds 57,58 , cultured endothelial cells rapidly lose the expression of vascular bed-specific markers [57][58][59][60] . HUVEC maintain expression of generalized endothelial function in cell culture, including shear stress responsiveness 5,61 . Hence, HUVEC are commonly used for mechanistic studies of endothelial cell behavior 62 . Moreover, HUVEC and HCAEC show similar in vitro responses to LSS stimulation 61 .

Conclusion
To summarize, in endothelial cells there is reciprocity between MAPK7 activity and the expression of EZH2. This reciprocity is regulated in part by a complex mechanism involving microRNAs and the regulation of phosphatase activity (Fig. 8). Dysregulation in the reciprocity between MAPK7 activation and EZH2 expression is associated to the induction of EndMT and the severity of coronary artery disease. These insights contribute to a better understanding of the molecular and epigenetic mechanisms that underlie endothelial homeostasis, the induction of EndMT during coronary artery disease and might represent a new target for therapy.

Materials and methods
Human coronary artery samples. Seventeen unique human coronary artery samples were obtained from autopsy specimens from 10 patients (age 59.1 ± 2.6 years, range 39-69) that died from an acute coronary episode at the Heart Institute (InCor), Sao Paulo, Brazil. Hypertension was present in 9 subjects, and diabetes in 6. Five individuals were active smokers. Coronary artery samples were stratified based on their respective intimamedia thickness (IMT) prior to further analyses. Stratified sample groups never contained more than one sample per patient. Next-of-kin gave informed consent and the investigation was performed according to institutional guidelines and approved by the Institutional CAPPesq Ethics Committee (InCor, Sao Paulo #SDC 3723/11/141 and #CAPPesq 482/11) and the Declaration of Helsinki 28 . All experimental protocols were approved by the Institutional CAPPesq Ethics Committee (InCor, Sao Paulo, Brazil). During necropsy each dissected coronary artery was fixed in neutral-buffered formalin with constant perfusion at a quasi-normal perfusion pressure before paraffin embedding.

Determination of intima-media thickness.
Four micron-thick sections were prepared from human coronary artery samples and deparaffinized using Xylol and rehydrated using a series of EtOH solutions of decreasing concentration. Samples were stained in Verhoeff 's solution (92 mM hematoxylin, 137 mM FeCl 3 , 27 mM KI, 4 mM I 2 in 55% EtOH) at room temperature for 1 h. Samples were differentiated in FeCl 3 (123 mM in dH 2 O) for 1 min and treated with Sodium Thiosulphate (316 mM in dH 2 O) at room temperature for 1 min. Samples were dehydrated using increasing concentrations of EtOH and cleared in 100% xylene. Samples were mounted in Permount resinous mounting medium. The intimal thickness was determined as the distance between the inner elastic lamina and the lumen, and the medial thickness was determined by measuring the www.nature.com/scientificreports/ distance between the inner elastic lamina and the outer elastic lamina at ten sites within one sample. Intimal/ Medial thickness was calculated by dividing the average intimal thickness by the average medial thickness 28 .

Endothelial cell culture and uniform laminar shear stress experiments. Human umbilical vein
endothelial cells (HUVEC, Lonza #C2519) were cultured in endothelial cell culture medium (ECM) as described previously 5,63 . EndMT was induced by the addition of 10 ng/ml TGFβ1 to the culture medium as described before 5,64 . For shear stress experiments, HUVEC (60 000 cells/cm 2 ) were seeded on 0.1% gelatin-coated µ-Slides I 0.4 Luer (Ibidi GmbH, Martinsried, Germany) and allowed to adhere under standard culture conditions overnight. Slides with a confluent endothelial cell monolayer were exposed to uniform laminar shear stress (20 dyne/ cm 2 ) for 24 h. Where indicated, 10 µM of the small molecule inhibitor of MAPK7 (BIX02189, SelleckChem, Munich, Germany) 14  A retroviral construct encoding the constitutively active rat MEK5-α1 (pBabePuro-MEK5D) and empty vector controls were kindly provided by Prof.dr. M. Schmidt (Dept. Dermatology, University Würzburg, Germany). Retroviral transduction of HUVEC was performed as detailed before 66 . In brief, virus-producing Phoenix cells were cultured until 70% confluency, after which basal medium was replaced by ECM after which viral supernatants were collected twice at 24 h intervals.

RNA isolation and transcript analysis.
Sections of whole arterial thickness were deparaffinated using xylol and rehydrated prior to homogenization in TRIzol (Invitrogen Corp, CA, USA). Cell cultures were lysed directly in TRIzol. RNA was isolated using the TRIzol reagent according to the manufacturer's protocol. RNA concentration and purity were assessed using UV spectrometry (Nanodrop 1000, Thermo Scientific MA, USA) and RNA integrity validated on 1% agarose gels. For gene expression analysis, cDNA synthesis was performed using RevertAid™ First Strand cDNA Synthesis Kit (Thermo Scientific, MA, USA), according to the manufacturer's protocol. For microRNA transcript analysis, 10 ng of total RNA was reversely transcribed using the ABI Taqman microRNA reverse transcription kit (#4366597, ThermoFisher Scientific) according to manufactures instructions using 1.0 µM microRNA-specific stemloop primers (Table 1). For all transcript analyses, the cDNA was amplified on a VIIA7 thermal cycling system (Applied Biosystems, Carlsbad, CA) in a reaction contain- 3'UTR binding assays. 3'UTR fragments were isolated, purified and cloned in the psiCKECK-2 reporter vector as described previously 64  COS7 cells were transfected with 100 ng UTR reporter plasmid and 50 pmol microRNA mimics as detailed above. 48 h post-transfection, luciferase activity was assayed using the DualGlo Luciferase assay system (Promega, Madison, WI) and recorded for 500 ms on a Luminoskan ASCENT (Thermo Scientific, Waltham, MA) according to manufacturer's instructions. Relative luciferase activity was calculated by dividing the luminescence from Renilla luciferase activity by the luminescence from firefly luciferase activity and normalized to control samples. chromatin-immunoprecipitation (ChIP) and assessment of histone modifications. Cells were harvested using accutase, pelleted and the chromatin crosslinked using 1% formaldehyde (37% F1268 Sigma-Aldrich) for 8 min. Crosslinking activity was quenched using 125 mM glycine (104201 Merck). Cell pellets were lysed on ice with SDS lysis buffer (1% SDS, 50 mM Tris HCl pH 8.0, 10 mM EDTA) supplemented with freshly added 100 mM protease inhibitor cocktail (Sigma Aldrich P8340) for 15 min. The chromatin was fragmented by Biorupter (Diagenode, Seraing, Belgium) with five cycles of (30' ON/OFF). The sonicated sample was centrifuged and chromatin containing supernatant was kept for further analysis. The chromatin was diluted 10 times with RIPA buffer (0.1% SDS, 0.1% Sodium deoxycholate, 1% Triton-X100, 1 mM EDTA, 10 mM Tris-HCl pH 7.5, 140 mM NaCl, 0.5 mM EGTA) supplemented with 100 mM protease inhibitor cocktail. Immunoprecipitation was performed by 4 µg H3K27Me3 antibody (Merk Millipore 07-449) or IgG control (Abcam ab46540) added to the 40uL Dynabeads Protein-A (Life technologies, 10002D) coated tubes. Subsequently, the chromatin of 0.8 × 10 6 cells was added to antibody bound beads and incubated overnight at 4 C while rotating. The beads Table 1. Primer sequences for microRNA expression analysis. Stem loop: GTC GTA TCC AGT GCA GGG TCC GAG GTA TTC GCA CTG GAT ACG ACT TCA GTTA  Sense: TGC GGT ACA GTA CTG TGA T   miR-141  Stem loop: GTC GTA TCC AGT GCA GGG TCC GAG GTA TTC GCA CTG GAT ACG ACC CAT CTT TAC  Sense: TGC GGT AAC ACT GTCTG   miR-200a  Stem loop: GTC GTA TCC AGT GCA GGG TCC GAG GTA TTC GCA CTG GAT ACG ACA CAT CGTT  Sense: TGC GGT AAC ACT GTC TGG T   miR-200b  Stem loop: GTC GTA TCC AGT GCA GGG TCC GAG GTA TTC GCA CTG GAT ACG ACT CAT CAT TAC  Sense: TGC GGT AAT ACT   Angiogenic sprouting assay. 10µL Matrigel (BD Corning, 356230) was added into the bottom compartment of µ slide Angiogenesis (81501, Ibidi GmbH, Martinsried, Germany) and incubated at 37 C, 5% CO 2 for 1 h. Cells were diluted to 2 × 10 6 cells/ml. 50 µl cell suspension was added on top compartment. After 6 h incubation at 37 C, 5% CO 2, light microscopy images were obtained, and complete octamer niches were counted by eye.  GGA GCT GGT GTG TTC TC  CAA AGC CGC CAT TTC ACC   −2.0 kb  1  GCG GTG ATG ATT AAC CCA AC  GTG GCC ACA GGT CAA GAA AT   −1.5 kb  1  GGT GAG AAC GCA ATG ACT GA  CTC CCA CTG CCA GGT TCA   −1.0 kb  1  TTG GAG GAG GAG ACT GGA AC  AGT TTT CTG GCA CCT TCC AC   −0.5 kb  1  GAC CAG CAG ACA CAC AAA CC  GAC CCC TCT CCC ATG CTG   TSS  1  TAC TGA GCT TCC CAG  www.nature.com/scientificreports/ Collagen contraction assay. Cells were dissociated using trypsin-EDTA, pelleted and suspended at a concentration of 22.5 × 10 6 cells/ml ECM. 45 µL cell suspension was added to a collagen solution (3.3 mg/ml rat tail collagen type I (#354236, BD, San Jose, CA), 100 mM Na 2 HPO 4 and 5 mg/ml NaHCO 3 ) of neutral pH. The cell/collagen mixture was immediately aliquoted into 50 µl droplets and allowed to polymerize at 37 C, 5% CO 2 for 30 min. Polymerized gels were released and 1 mL of ECM was added. At time points t = 0 h and t = 24 h, gels were visualized using a regular flatbed scanner and the gel surface area quantified using with ImageJ (NIH). Gel contraction was calculated as the relative reduction in gel surface area at 24 h.
Data representation of statistical analyses. Data are expressed as mean ± s.e.m. from at least three independent experiments. Where the mean of two groups were compared, p-values were calculated using student t-tests. Otherwise, p-values were calculated using the one-way analysis of variance (ANOVA) followed by Bonferroni's post-hoc comparisons tests using GraphPad Prism 9 (GraphPad Software, La Jolla, CA, USA). P < 0.05 was considered statistically significant.

Data availability
All data generated or analyzed during this study are included in this published article. Materials, data, and associated protocols are available from the corresponding author on reasonable request without preconditions. www.nature.com/scientificreports/